The role of unprotected and privately protected areas for ocelot conservation: densities in Colombia and Brazil

Abstract

The ocelot (Leopardus pardalis) is a widely distributed, medium-sized felid in the Americas with declining population size. We estimated ocelot densities and home ranges in one agricultural area in the Magdalena River valley in Colombia, a private reserve and cattle ranch in the Colombian Llanos, and a private reserve in the Serra do Amolar in the Brazilian Pantanal. We used camera trapping (39–52 stations) and spatially explicit capture-recapture (SECR) models. Density estimates (individuals/100 km²) were 11.0 ± 2.7 (SE) in the Magdalena River valley; 13.2 ± 3.2 (SE) in the Llanos, and 10.3 ± 2.9 (SE) in the Serra do Amolar. Overall, despite an impact of agriculture and human disturbance, our results highlight the importance of unprotected areas and privately protected ranching areas for ocelot conservation. As agriculture continues to expand across the tropics causing habitat loss, and negatively affecting ocelot densities, we recommend land use planning and best agricultural practices to maintain natural habitats, thereby limiting human impacts on ocelot conservation.

El ocelote (Leopardus pardalis) es un felino de tamaño mediano ampliamente distribuido en las Américas con un tamaño de población descendente. Estimamos las densidades del ocelote y sus áreas de acción en una zona agrícola en el valle del río Magdalena en Colombia, una reserva privada y rancho ganadero en los Llanos colombianos, y una reserva privada en la Serra do Amolar en el Pantanal brasileño. Usamos cámaras trampa (39–52 estaciones) y modelos de captura-recaptura espacialmente explícitos (SECR). Las densidades estimadas (individuos/100 km²) fueron de 11.0 ± 2.7 (SE) en el valle del río Magdalena, 13.2 ± 3.2 (SE) en los Llanos, y 10.3 ± 2.9 (SE) en la Serra do Amolar. En general, a pesar del impacto de la agricultura y la perturbación humana, nuestros resultados destacan la importancia de las áreas no protegidas y las áreas ganaderas privadas protegidas en la conservación del ocelote. A medida que la agricultura continúa expandiéndose en los trópicos causando pérdida de hábitat y afectando negativamente las densidades de ocelotes, recomendamos la planificación del uso de la tierra y mejores prácticas agrícolas para mantener los hábitats naturales en dichas áreas, limitando el impacto antrópico sobre la conservación del ocelote.

camera trap, carnivore, cattle ranching, felid, Llanos, Magdalena river, oil palm, Pantanal, spatially explicit capture-recapture

cámara trampa, captura-recaptura espacialmente explícita, carnívoro, ganadería, félidos, Llanos, palma de aceite, Pantanal, río Magdalena

The ocelot (Leopardus pardalis) is a medium-sized spotted felid and the third largest Neotropical felid (Sunquist and Sunquist 2002), with a distribution ranging from Southwestern United States to Argentina (Paviolo et al. 2015). Ocelots were heavily hunted for their fur in the 1960s and 1970s (Payán and Trujillo 2006; Salvador and Espinosa 2015) and were listed as Vulnerable on the International Union for the Conservation of Nature Red List from 1982 to 1990. Today, due to the enforcement of the Convention on International Trade in Endangered Species ban on their hunting and their wide distribution, they are considered Least Concern (Paviolo et al. 2015); yet their population is decreasing due to habitat loss and fragmentation, as well as persecution related to small livestock depredation and conflict with humans (Paviolo et al. 2015). However, ocelot density estimates and population size data are relatively scarce outside national parks, and especially in ever-expanding agricultural landscapes.

Mesopredators can have an indirect effect on forest structure and composition. In the case of ocelots, they play a key role in keeping small- and medium-sized prey populations, such as agoutis and spiny rats, in check (Emmons 1987; Salvador and Espinosa 2015). This is important because rodents are seed consumers but also dispersers, especially when buried seeds are forgotten or their owners preyed upon. Thus, predation indirectly helps to guarantee healthy forest regeneration (DeMattia et al. 2004; Roemer et al. 2009). Ocelots are considered ecologically plastic, inhabiting natural areas from dense forests to open savannas, and can survive in modified areas such as agricultural landscapes as long as some proportion of forest cover remain in the landscape (Diaz-Pulido and Payán 2011; Massara et al. 2015; Rocha et al. 2016).

Ninety percent of carnivore ranges fall outside protected areas (Di Minin et al. 2016), thus unprotected areas are key for connectivity and survival of wide ranging species like felids and need to be integrated into conservation strategies (Rabinowitz and Zeller 2010; Di Minin et al. 2016; Boron et al. 2020). Furthermore, with ever-increasing agricultural expansion in the tropics (Pendrill et al. 2019), it is crucial to investigate species ecology in these landscapes to guide conservation management. The latter is particularly important as there is a danger of ocelot extirpation in highly modified and fragmented lands (Janečka et al. 2011). Many of these areas are private lands, hence working with landowners across ocelot ranges to ensure forest preservation will be crucial to ocelot conservation (Haines et al. 2006a; Paolino et al. 2018).

Estimates of population densities are crucial for conservation and management strategies. They are important indicators of population health across the species range, can help refine distribution patterns, and they provide insights on species tolerance levels to human-modified landscapes (Boron et al. 2016a). Densities can also be compared across space and time to reveal population trends and detect local declines and threats (Massara et al. 2015; Rocha et al. 2016; Satter et al. 2019a). Ocelot densities vary greatly (Noss et al. 2012; Salvador and Espinosa 2015; Satter et al. 2019a) and tend to increase with higher rainfall and lower latitude (Di Bitetti et al. 2008). Home ranges also vary, with male home ranges generally larger than those of females (Salvador and Espinosa 2015; Rocha et al. 2016; Satter et al. 2019b).

Camera trapping combined with spatially explicit capture-recapture (SECR) has been the most robust approach to estimate densities of elusive and individually marked carnivores like ocelots (Royle and Young 2008; Sollmann et al. 2011; Satter et al. 2019b). In this study, we use camera trap data and SECR to estimate ocelot densities in three South American sites: an agricultural landscape in the Magdalena river valley in Colombia (1), an extensive cattle ranch and private reserve in the Colombian Llanos (2), and contiguous private protected areas in the Serra de Amolar in the Brazilian Pantanal (3). The Colombian Llanos and Magdalena areas are used for cattle ranching, which is an established land use throughout Latin America (Grau and Aide 2008). The Magdalena area additionally has oil palm cultivation, an emerging land use in the region (Ocampo-Peñuela et al. 2018). The Serra do Amolar is well conserved with no agriculture. Both Colombia and Brazil are key countries for the ocelot’s long-term conservation, despite scarce information, especially in Colombia where only one ocelot study has been published (Diaz-Pulido and Payán 2011).

We hypothesize that increased disturbance from agriculture and livestock will impact ocelot density negatively. Therefore, we expect that ocelot density in the Magdalena region and the Llanos would be lower than values recorded in similar biomes due to the impact of agriculture and cattle ranching, while density in the Serra de Amolar, Pantanal would be indicative of what the biome can sustain. However, there are no previous SECR density estimates to compare with the Llanos and the Pantanal. Our data advance understanding of ocelot ecology and populations highlighting the potential role of production, unprotected, and private conservation areas for ocelot conservation.

Material and Methods

Study areas

We conducted the study at two sites in Colombia and one in Brazil (Fig. 1). Site I is located in the Magdalena River inter-Andean valley (7° 22′ 30.7″N, −73° 53′ 3.0″E to 7° 32′ 25.6″N, −73° 42′ 42.4″E) in the Department of Santander, Colombia. The region has a tropical climate with mean annual temperature of 27 °C and 2,100 to 2,600 mm of annual rainfall (IDEAM et al. 2007) concentrated in April–May and between October and December. It is naturally covered by humid tropical forests and wetlands (IDEAM 2014), but it has largely been converted to cattle ranches and oil palm plantations. However, the region still hosts endemic and endangered species, and it provides important connectivity between protected areas for several species (Payan-Garrido et al. 2013). Main land cover types are pasture (35%), wetlands (20%), oil palm plantations (19%), secondary forest (12%), water (10%), bare ground (3%), and urban areas (<1%) (Etter and van Wyngaarden 2000; Castiblanco et al. 2013; Boron et al. 2020).

Fig. 1.

Map of the study sites with camera locations. Site I is part of the Magdalena River valley in Colombia, Site II is located in the Colombian Llanos and Orinoco River basin, and Site III is in the Serra do Amolar, Brazilian Pantanal.

Site II is Hato la Aurora, a private nature reserve and cattle ranch in the Orinoco River basin in the Llanos region and in the Department of Casanare, Colombia (5° 57′ 18.8″N, −71° 29′ 0.1″E to 6° 4′ 52.6″N, −71° 17′ 51.4″E). Mean annual temperature is 27 °C and average rainfall is between 1,000 and 3,000 mm concentrated between April and November (IDEAM et al. 2007). This area is a highly biodiverse seasonally-flooded tropical savannah dissected by riparian forests, and the dominant land use is extensive cattle ranching with introduced grasses (IDEAM 2014). The main land covers are primary and secondary riparian forest (38%), native and introduced grasslands (31%), scrubland (30%) and water (1%) (IDEAM 2014).

Site III is in the Serra do Amolar in the Brazilian Pantanal ecoregion, located in the states of Mato Grosso and Mato Grosso do Sul, Brazil (−17° 57′ 40.7″N, −57° 28′ 48.3″E to −18° 11′ 53.8″N, −57° 23′ 20.1″E). This area ranges in altitude from ~80 m.a.s.l. on the banks of the Paraguay River to the highest peak at 1,000 m (Porfirio et al. 2014). Mean temperature is 25 °C (Rohli and Vega 2008; Fernandes et al. 2010) and the average annual rainfall is 1,400 mm, with variation between 800 and 1,600 mm, concentrated during the rainy season from January to June (ANA Agência Nacional das Águas et al. 2005). This area is characterized by a mosaic of forests and riparian forest along the Paraguay River (59%) and associated waterways and wetlands (35%), dry and humid savannahs (6%) (Porfirio et al. 2014; IBGE Instituto Brasileiro de Geografia e Estatística 2020). The areas surrounding the study site in the Serra do Amolar present a high deforestation index with extensive agricultural areas and urban regions (Casagrande and Santos-Filho 2019), though the study area consists of a series of contiguous private natural heritage reserves with no production activity and good forest cover.

Camera trapping

We conducted camera trapping surveys between the end of April and early August 2014 at Site I (both dry and rainy seasons), in April–May 2014 at Site II (rainy season), and December 2019–February 2020 at Site III (end of dry season and beginning of rainy season) without the use of any bait or lure. Despite season transitioning in Sites I and III, habitat conditions can be considered stable with no flooding occurring in and around camera grids. Furthermore, seasonal changes do not have a significant effect on ocelot relative abundance (Negrões et al. 2011) and occupancy (Massara et al. 2015).

We used a blocked design (i.e. two adjacent and subsequent blocks) at Site I, and a continuous design at Site II and Site III. Our camera grid consisted of 47 stations across 154.8 km² (Minimum Convex Polygon, MCP) at Site I, 53 stations across 151.3 km² at Site II, and 39 stations across 190 km² at Site III. The MCPs are of appropriate size for ocelot density studies (Massara et al. 2015; Rocha et al. 2016; Satter et al. 2019a). We used Cuddeback Attack (model: 1149) and Ambush (model: 1170) camera traps at Site I, Panthera cameras (Series 3 and 4) at Site II and Cuddeback (model: 1279) and Bushnell (model: 119876) camera traps at Site III. We programmed them for continuous operation of 24 h with a 30-s interval between photos. We checked cameras every 30 days where necessary to change batteries and retrieve memory cards.

The studies were originally designed to estimate jaguar densities (Boron et al. 2016a) but we kept a conservative distance between camera stations in order to obtain robust data for the wider mammal community. At all three sites we placed paired camera stations in a grid with a spacing of 1.6 ± 0.2 km, which covered all habitats of the regions, and at a height of 0.40 m from the ground level. Paired cameras ensure photographing both flanks of each passing individual, enabling individual identification. Where possible we placed cameras on trails to maximize carnivore capture probability. The distance between stations is consistent with previous ocelot density studies (Salvador and Espinosa 2015; Rocha et al. 2016; Satter et al. 2019a) and is appropriate when considering ocelot home ranges estimates (Salvador and Espinosa 2015; Satter et al. 2019a). Finally, we limited the surveys to less than 100 days. Overall, our study design complies with capture recapture model assumptions at all three sites, that is populations can be considered closed and stable due to the short sampling period, and all individuals should have at least some probability of being captured (Otis et al. 1978; White 1982).

Data analysis

We identified ocelot individuals from their spot patterns and their sex based on photographic evidence of external genitalia. We then estimated density fitting SECR models in a maximum likelihood framework (Borchers and Efford 2008; Efford et al. 2009) and using the package “secr” in R (Efford 2020). SERC models use individual spatial locations to define their activity centers (or home range centers) and then estimate density of these centers across a polygon that includes the camera grid (Efford 2004; Royle and Young 2008).

Model assumptions are that home ranges are circular and constant during the survey, individual activity centers are randomly distributed, and the encounter rate of an individual with a trap decreases with increasing distance from the activity center following a predefined function (Efford 2004; Royle and Young 2008). We deployed the half-normal detection function where the probability of capture (P) of an individual (i) decreases with distance (d) from the activity center as: P_ij= g₀exp(−d_ij²/2σ ²), where g₀ is the probability of capture when the trap j is located exactly at the center of the home range, and sigma (σ) is a spatial parameter related to home range size (Efford 2004). As in other camera trap studies, we used the binomial encounter model (or Bernoulli model) where an individual can be recorded at different camera stations during each sampling occasion but only once at each station (Royle et al. 2009; Noss et al. 2013). Since felid populations have unequal ranging patterns and behaviors between sexes (Massara et al. 2015; Satter et al. 2019a), we allowed both parameters g₀ and σ to vary with sex of the individuals (Sollmann et al. 2011; Tobler et al. 2013) and compared four models using the Akaike information criterion (AIC; Burnham and Anderson 2002): “SECR.0” (null model), “SECR.g0” (g₀ varies between males and females), “SECR.σ” (σ varies between males and females), and “SECR. sex” (both g₀ and σ vary between sexes).

Results

We recorded 21 adult ocelot individuals at Site I (54 events), 25 at Site II (44 events) and 18 (42 events) at Site III. We were able to determine sex in 53 individuals (Table 1). With exception of Site 1, males were recaptured more frequently than females. Several cameras stations were visited by different individuals, with the most visited station occurring at Site II with five individuals: three males and two females.

Table 1.

Survey features for Site I, Site II, and Site III. N = number of individuals.

	Site I	Site II	Site III
Location	Magdalena River valley	Orinoco River basin	Serra do Amolar
Survey period	April–August 2014	April–May 2014	December–February 2020
Camera trap stations	47	52	39
Minimum Convex Camera polygon (km2)	154.8	151.3	190.0
Trap nights	2,251	2,457	1,670
Total N recorded (capture events)	21 (54)	25 (44)	18 (42)
N females (capture events)	6 (17)	15 (23)	4 (10)
N females captured at different stations	1	3	3
N males (capture events)	11 (31)	10 (21)	7 (21)
N males captured at different stations	5	3	2
N unknown (capture events)	4 (6)	0	7 (11)

The best model for all three sites was the null model (Table 2). However, for Site II, the model SECR.g0 also had strong support (ΔAICc < 2). The estimated parameters associated with probability of capture (g₀), home range (σ), and densities (D) across the three sites are reported in Table 3. Density estimates ranged from 10.2 to 13.2 individuals/100 km² and did not vary significantly among the three sites, displaying overlapping 95% confidence intervals (CIs; Table 3). We estimated home range sizes as 35.89 km² at Site I, 38.00 km² at Site II, and 25.89 km² at Site III.

Table 2.

Model selection parameters for spatially explicit capture recapture (SECR) models at Site I, Magdalena River valley, Colombia; Site II, Orinoco River basin, Colombia; and Site III, Serra do Amolar, Brazil.

	Site I				Site II				Site III
	AICc	ΔAICc	W	K	AICc	ΔAICc	W	K	AIC	ΔAICc	W	K
SECR.0	307.93	0.00	0.67	4	264.02	0.00	0.53	4	200.12	0.00	0.73	4
SECR.g0	310.80	2.87	0.16	5	265.28	1.26	0.28	5	203.31	3.19	0.15	5
SECR.σ	311.00	3.07	0.14	5	266.89	2.87	0.13	5	204.03	3.91	0.11	5
SECR.sex	314.73	6.80	0.02	6	268.47	4.45	0.06	6	207.80	7.68	0.02	6

AIC = akaike information criterion; ΔAIC = difference in AIC values between each model and the model with the lowest AIC; W = AIC model weights; K = number of model parameters. SECR.0: null model. g₀ = probability of capture at the home range centre, σ = spatial parameter related to home range size; SECR.0: null model; SECR.g0: g₀ varies between males and females; SECR.σ: σ varies between males and females; SECR.sex: both g₀ and σ vary between males and females.

Table 3.

Density and parameters estimated by the best spatially explicit capture recapture (SECR.0) models at Site I, Magdalena River valley, Colombia; Site II, Orinoco River basin, Colombia; and Site III, Serra do Amolar, Brazil.

	Site I				Site II				Site III
	Value	SE	95% LCI	95% UCI	Value	SE	95% LCI	95% UCI	Value	SE	95% LCI	95% UCI
g0	0.01	0.00	0.01	0.02	0.01	0.00	0.00	0.01	0.03	0.01	0.02	0.05
σ (km)	1.38	0.15	1.11	1.71	1.42	0.20	1.08	1.87	1.17	0.16	0.89	1.53
D (N/100 km2)	10.97	2.72	6.80	17.70	13.24	3.18	8.33	21.05	10.27	2.86	6.01	17.55

Discussion

Ocelots play important ecological roles; however, their populations are decreasing (Paviolo et al. 2015). This is the first study that estimates ocelot densities in unprotected and privately protected agricultural areas in Colombia and Brazil. Estimating density in modified landscapes promotes understanding of how resilient ocelots are and how they can persist in anthropogenic areas. With ongoing habitat loss in the tropics (Pendrill et al. 2019) such information is vital to inform conservation strategies, evaluate status for red lists, and construct survival thresholds. Our results revealed densities of 11.0 ± 2.7 (SE) individuals/100 km² in the Magdalena river valley, 13.2 ± 3.2 (SE) in the Colombian Llanos and 10.3 ± 2.9 (SE) in the Serra do Amolar.

Ocelot densities estimated with SECR models are still scarce; however, we base our discussion exclusively on those estimates to avoid making comparison with previous results based on nonspatial methods. Such direct comparisons would not be meaningful as studies prove that there is variation between spatial and nonspatial estimates (Noss et al. 2012). All density estimates we refer to are expressed as number of individuals per 100 km². At a continental scale, ocelot densities tend to increase with rainfall, primary productivity, and at lower latitudes, which also reflect higher prey availability (Di Bitetti et al. 2008). They are usually higher in tropical wet forests such as Amanã Reserve, Brazilian Amazon, with estimates of 24.8 ± 6.3 (SE) (Rocha et al. 2016) and in large protected areas, for example 51.7 ± 10.4 (SE) in the Kaa-Yaa National Park in Bolivia (Noss et al. 2012). On Barro Colorado Island in Panama, ocelot densities reach 159.0 ± 46.4 (SE; Rodgers et al. 2014), but here they have experienced competitive release due to absence of jaguar (Panthera onca; Moreno et al. 2006). In drier and more open habitats, densities are as low as 3.2 ± 0.5 (SE) in Semi-arid Caatinga in Brazil (Penido et al. 2016), and 0.9 (95% CI: 0.6–1.2) in a Native upland tropical pine forest in Belize (Satter et al. 2019a).

Site I is a tropical lowland forest biome with wetlands and seasonal flooded zones (IDEAM 2014). Our density estimate (11.0 ± 2.7 SE; 95% CI: 6.8–17.7) constitutes the first estimate of densities of ocelots in an inter-Andean valley. It was lower than in protected areas in Amazonian tropical lowland forests, where Rocha et al. (2016) reported an average density of 24.8 ± 6.3 (SE) from three estimates for the Brazilian Amazon. The 95% CIs of the three estimates were between 10.9 and 48.6. The extensive habitat loss that occurred in the Magdalena region due to oil palm plantations and cattle ranching (Castiblanco et al. 2013) could explain our lower ocelot densities albeit overlapping CIs. The latter does not necessarily prove a lack of differences but perhaps the precision of density estimates does not allow detection of such fine differences, which is expected considering the sparse data obtained in felid surveys (Payán 2013). Site I contains secondary forest fragments that are limited to only 13% of the study area, while anthropogenic land covers amount to almost 60% (Boron et al. 2020). However, our density estimates are similar to tropical forests in Belize, where densities vary from 9.3 (95% CI: 4.5–19.1) to 13.0 (95% CI: 10.8–15.2) (Satter et al. 2019a). Forests and well-conserved areas tend to be smaller in size in Central America than South America, which could explain overall lower densities (Satter et al. 2019a).

The Colombian Llanos (Site II) can be considered comparable to the Pantanal in Brazil, where Site III is located. However, annual flooding lasts longer than in the Pantanal which limits the possibility of human land use and human impact on the ecosystem (Boron et al. 2016a). In addition, colonization of the Llanos started 200 years earlier than in the Pantanal, resulting in a larger human population and greater hunting pressure. Furthermore, our site in the Llanos had cattle ranching, while there was no use of land at Site III. Due to all of these reasons, we expected a lower density at Site II. However, density estimates were comparable between sites (Llanos: 13.2 ± 3.2 SE and Serra do Amolar: 10.3 ± 2.9 SE). This may be due to the study area at Site II being a well-conserved cattle ranch and private reserve, where despite productive activity, forests and natural habitat are preserved and hunting is forbidden. It is not possible to compare our density estimates at these two sites with those previously documented in the Llanos and the Pantanal, since previous estimates are based on nonspatial methods. The values we report are lower than estimates in tropical lowland forests which can be explained by lower productivity and forest cover in our study sites compared with the Amazon (Di Bitetti et al. 2008). Ocelot preference for forest is well documented (Michalski and Peres 2005; Paolino et al. 2018; Boron et al. 2019; Wang et al. 2019). Our study sites in the Llanos and the Pantanal had 38% and 59% forest cover respectively, while the remaining proportion of the landscapes is composed of open habitats such as grasslands and water bodies.

Ocelot home ranges tend to vary according to prey availability, sex, season, and region (Dillon and Kelly 2008; Paviolo et al. 2015). They range between 1 and 43 km² but few estimates are available (Emmons 1988; Silver et al. 2004; Dillon and Kelly 2008; Maffei and Noss 2008; Paviolo et al. 2015; Azevedo et al. 2019). The values we estimated (35.9 km² at Site I, 38.0 km² at Site II, and 25.9 km² at Site III) are on the higher side of the range, especially for Site I and Site II, possibly indicating slightly less favorable conditions and stronger disturbance. For example, home ranges are usually larger when prey is more limited (Di Bitetti et al. 2006; Dillon and Kelly 2008; Azevedo et al. 2019). Large home ranges at Site II may also be due to the need to range between different riparian forests to find sufficient prey. However, our estimates should be treated with caution. SECR home range estimates, derived from σ, are based on the length of individual movements between stations (Efford et al. 2009). Shorter movements are not recorded, potentially leading to an overestimation of the home ranges. In all three sites, we detected only a limited number of individuals (24–28%) in more than one camera station, thus we recommend decreasing the distance between stations in future surveys.

Overall, male ocelots generally have larger home ranges than females, and male individuals can overlap with more than one female (Dillon and Kelly 2008). This translates to higher detectability of males in camera trap surveys (Satter et al. 2019a). Given these differences between males and female ocelots, we expected that the best models would have been those that took into consideration sex variation for g₀ and σ. This was not the case and it could be because at both Site I and Site III we were not able to identify the sex of several individuals. At Site II, where we had greater data availability for both sexes since we were able to sex all individuals, the model SECR.g0 also had good support.

Our study provides important data on ocelot densities in an agricultural, unprotected area (Site I), a cattle ranch and private reserve (Site II), and in private protected areas with no production activities (Site III), advancing understanding of the ecology of this species in different environments. Our a priori expectation that lower densities would be found in more altered sites is only partially supported. Ocelots showed lower densities at Site I compared with well-conserved areas in similar ecosystem types, which highlights that they are likely affected by agriculture and human disturbance. However, differences were not significant (overlapping CIs), and there was no evident impact of cattle ranching at Site II, denoting a certain amount of resilience to anthropogenic habitat alterations.

Furthermore, the densities we report here are still higher than in several other parts of the ocelot range, highlighting the importance of unprotected and private conservation areas for wide ranging ocelot conservation. Their survival in unprotected lands might well hold the key to their long-term survival, enabling connectivity, and gene flow among better conserved core areas. The extent of disturbance that populations can endure is still unclear, and ocelots can certainly become locally extinct in highly fragmented areas (Janečka et al. 2011). As agriculture continues to expand across the tropics, often at the expense of forests (Pendrill et al. 2019), it is crucial to ensure that large forested areas are maintained (Horne et al. 2009) and that riparian forests are respected (Paolino et al. 2018).

Ocelots and associated biodiversity will benefit from stronger regulatory frameworks that facilitate land use and infrastructure planning, taking into account priority forest areas with high conservation value to steer further agricultural expansion on already modified areas (Haines et al. 2006b; Garcia-Ulloa et al. 2012; Boron et al. 2016b). Agricultural areas are private lands, and since cattle ranching is the main land use in South America (De Sy et al. 2015) it is crucial to work with landowners (Haines et al. 2006a), and especially ranchers, to ensure forest conservation and limited hunting on private properties (Haines et al. 2006a; Paolino et al. 2018). Ocelot populations can survive on extensive livestock ranches such as the one documented here and elsewhere (Haines et al. 2006a; Diaz-Pulido and Payán 2011). The same applies for private reserves like Site III, which have a demonstrated value for ocelot and other biodiversity conservation (Negrões et al. 2011). Incentive-based approaches (e.g., tax breaks, subsidized credits, and premium prices), and nature-based livelihoods like ecotourism, can encourage landowners to preserve natural areas (Lambin et al. 2014; Boron et al. 2016b). Finally, more research is needed to understand additional factors that may be influencing ocelot densities and home ranges, such as prey availability, seasonality, presence of other carnivores, and different anthropogenic pressures.

Acknowledgments

We would like to thank C. Diaz for elaborating the study map and the Panthera office staff for helping with general logistics. Thanks to A. Quiñones Guerrero, J. Murillo, and Cabildo Verde in the Magdalena, the Instituto Homem Pantaneiro in the Pantanal, the entire Barragán family in the Llanos, and R. Ortiz for their help during fieldwork. Thanks also to our field guides, to the landowners and workers for allowing us to work in their properties.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

Funding for this study was provided by Panthera; Liz Claiborne and Art Orternberg Foundation, Jaguar Research Grant; Interconexión Eléctrica S.A. (ISA); The Rufford Foundation, Small Grant #14968-1; The Explorers Club, Exploration Fund grant; and Idea Wild.